A NEW WAY TO LOOK AT TRAFFIC INCIDENTS

Loop detectors and other sensors as used today in traffic management systems often provide less than adequate detection of traffic incidents. The primary reason is that the sensors / systems attempt to detect congestion that infers the presence of an incident rather than directly detecting the same. The purpose of this paper is to present the results of research into a revolutionary approach to traffic monitoring with rapid incident detection as a focus. The methodologies are based upon image processing techniques which directly detect traffic incidents. This method may be used within an ATMS infrastructure to IMMEDIATELY, AUTOMATICALLY, and RELIABLY produce an incident alarm in the form of a video image at an operator's station, in daylight or dark and good weather or poor, showing the scene in which an incident has been detected with the detected anomaly surrounded by a colored box. Further work in this area can lead to automatic classification of the severity of a detected incident or of the potential for further traffic hazards and injuries, etc. Certain traffic hazards which are accidents "about to happen" are also automatically detectable using these techniques, providing operators with opportunities to avoid potential accidents by rapid intervention.